A Comparison of the Quality of Questions in a Face-to-face and a Web Survey

Up until now, most surveys have used face-to-face interviews, postal mail, or the telephone to collect data. However, today the difficulties of carrying out surveys at reasonable costs have increased. Simultaneously, new opportunities have appeared. In particular, web surveys, which are usually cheaper, offer more flexibility, and can reach a large population in a short time, are becoming very attractive. Nevertheless, different modes of data collection may lead to different coverage, sampling, nonresponse, and measurement errors. We focus on the last since different modes have different properties, just because the question is asked in a different mode, a difference in responses may appear. For instance, Krosnick (1991) shows that varying levels of social desirability and satisficing biases exist depending on the mode of data collection used. This can be related to the presence of an interviewer in some modes but not in others. As a result, in order to compare data collected with different modes (across time, across countries, across groups), we first need to study the impact of modes on several parameters.